Docteract, a portmanteau of "Document" and "Interact," is a chatbot that revolutionizes the way you interact with documents. Upload a document in txt, pdf, or docx format, and engage in a question-and-answer session with our intelligent bot to efficiently extract the information you need.
- Document Upload: Supports txt, pdf, and docx formats for document upload.
- Intelligent Q&A: Uses a chatbot to provide answers from your uploaded document.
- User-Friendly Interface: Seamless and intuitive for easy navigation and interaction.
- Privacy-Focused: Utilizes an open-source language model, ensuring enhanced privacy and security for user data and interactions.
- Frontend: Developed using React with Vite and TypeScript for a responsive user interface.
- Backend: Python-based, ensuring reliable processing of user interactions and document handling.
- Language Model: Utilizes the open-source language model "teknium/OpenHermes-2p5-Mistral-7B" for chatbot functionality.
- Deployment: The model is deployed on Together AI, and the backend is hosted on AWS LightSail with the frontend on Vercel.
- Language Model Scope: The chatbot, powered by a 7 billion parameter model, may not be as comprehensive as models with larger parameters like GPT-3.5.
- Scalability: Currently using free tiers for hosting; full scalability for larger user loads is not yet implemented.
- Language Support: Optimized primarily for English language documents.
- Document Content Limitations: The application is currently unable to process documents with image-based content or complex formats, such as tables. Users are recommended to upload documents containing pure text for optimal interaction and results.
Docteract is released under the MIT License, permitting free usage and distribution with attribution, making it suitable for open-source projects.